Problem

Modern users take large numbers of screenshots every day, including receipts, timetables, notes, social media posts, and important information. Over time, these screenshots become disorganized and difficult to search through, leading to wasted time and cluttered galleries. Most devices currently rely on manual folder creation and sorting, which is inconvenient and rarely maintained by users consistently.

Solution

Sortable is an AI-powered downloadable application that automatically organizes screenshots into smart folders based on their content. Screenshots are base64-encoded and sent to Llama 4 Scout as a vision task. In existing mode it picks folders you already have; in LLM mode it can create new categories to store the screenshot. Using image and text recognition, the app detects categories such as finance, study, travel, memes, or work-related screenshots and sorts them instantly, creating a cleaner and more efficient experience.

Two classification modes:

• Existing Folders Mode: App walks the current screenshot directory up to 2 levels deep and feeds folder names to the AI model. Prompt will instruct the AI model to pick exactly 1 folder. Invalid responses fall back to root folder.
• LLM creates folders Mode: The model is given the current known folder list from config and prompted to either match an existing category or invent a new short folder label. New names are saved back to config so the model can see them and use them on the next run.

Tech

Sortable is built using Python and the Groq API to power AI-based screenshot analysis and categorization. The application processes screenshots, extracts relevant text and visual information, identifies patterns and keywords, and automatically sorts files into smart folders through a lightweight desktop interface.

The Stack: App Shell: Python + Rumps (Python framework for Mac OS Menu Bar) Native UI: AppKit, PyObjC File watching: Watchdog (for filesystem events) AI model: Groq API with Meta Llama 4 Scout (embedded API key for demo purposes) Packaging: PyInstaller (bundling into a standalone .app)

Images are sent to Groq API and not stored.

Try it out: (Mac OS Only MVP)

Download and follow the instructions (also below) on the latest release on our GitHub repo. Groq API key is embedded into the code; please notify @king_dog_. on Discord if there are any problems :)

Installation instructions:

Download Sortable.zip in the GitHub release Unzip → drag Sortable.app to Applications Right-click → Open if macOS blocks it on first launch In the setup wizard, pick a screenshot folder, and select your preferred mode Done! Sortable is ready to use.

Built With

• appkit
• groq
• json
• pyinstaller
• pyobjc
• python
• rumps
• watchdog